UIUC Geology Professor Bruce Hannon tells his night class in Advanced Ecological Modeling, "Computers are an aid to the mind just as telescopes and microscopes are aids to the eye." A mixture of students and scientific researchers who hold daytime jobs, the class is building a 200,000-cell model of a 20 mile x 30 mile desert steppe ecosystem on an army range near Yakima, WA--home to the sage grouse.
The class is divided into small groups, each of which works on a sector of the larger model that can be run independently of the whole. Group members build and test their sectors during the week, then work with the class to assemble the full cell model.
Tonight members of each group take turns sharing their progress as a model is projected on a large screen. James Westervelt, staff member at the nearby U. S. Army Construction Engineering Research Laboratories (CERL) and developer of the Geographical Resources Analysis Support System (GRASS) used in the model and who teaches with Hannon, manipulates the projection while Hannon conducts the discussion.
The Soils Group is describing how litter affects moisture retention in the soil. The Plant Group wants to know if the litter creates enough organic matter when broken down to affect plant growth. The Soils Group does not know, but everyone agrees that the Fires Group needs the numbers on litter.
Later, the Grouse Group asks the Plant Group how much food will be available for the birds. As sections of the model are being projected, graphs are being shown, run, and changed-- inching the class closer to an accurate generic cell model of the ecosystem. As a whole, the class wants to build the largest dynamic model ever run and develop a method for building a global ecological model.
"STELLA lets us bring people together who are experts in science, but who may have little or no programming and math experience," Hannon explains. "The basics of using STELLA are so easy to learn that expertise doesn't matter.
"We take these scientists and their students and build a generic ecological model. Once we all agree on the correct dynamics in terms of the way plants and animals interact, we turn the information over to Albert [Albert Cheng, NCSA research programmer] and specify to him the precise cell size and time-step size. To make the model cell-specific, we describe the rules governing the migration between cells, in this case the movement of sage grouse (an endangered species) and the movement of troops and army vehicles. Albert translates the generic model into Fortran, parameterizes it for each cell, and defines the migration. Then we can run groups of cells in parallel on the CM-5. We'll do this for the first time in this class.
"We are introducing another new concept in ecological modeling: the use of STELLA, the CM-5, and the GRASS geographical information system together. We're running fundamentally different programs together in the same model. At this point in the class we still have to add the parameterization and migration rules.
"Using this method, we calculate that we can run a very large STELLA model representing the grouse at various life stages, different kinds of plants and predators, soil type and moisture, weather--all the many physical variables--and also introduce the necessary human activities like tank and troop maneuvers on the army range. Each cell could contain 100 to 200 variables. The basic idea is to test the sensitivity of grouse populations to changes in the troops' training schedule."
Much remains to be done. The success of running the final full-scale model rests with Cheng. So far he has successfully run a simple 10-variable STELLA model on the CM-5, and he expects to run a full model by the end of the semester. Hannon wants a seamless connection between the Quadra 700 and the CM-5. Cheng, who describes the present set-up as "primitive," is building an automated translation engine that would convert STELLA input into Fortran for running on the CM-5. If it works, he would not have to manually translate each model change.
After the semester's end, Cheng continues working full time to get the desert steppe ecosystem model running on the CM-5. As Hannon says, "That requires a great deal of attention including some additional programming to connect all 200,000 cells with human activity on the desert and on the movement of sage grouse. Westervelt and Cheng are putting in prodigious amounts of effort. The project would simply not be 'com-pleteable' without them. We are solving a uniquely hard problem with some of the best in the business. We want to have a result by an invitational mini-conference we are conducting in March."
Cheng explains that what he, Hannon, and Westervelt are doing is similar to the dynamic ecosystem simulation modeling developed by Thomas Maxwell for Robert Con-stanza, professor of ecology at the University of Maryland and director of the Maryland International Institute for Ecological Economics, in his work on regional models (see access, July-September 1992). Maxwell's translator model was written in C to run on a Macintosh with transputers; Cheng is translating that model into Fortran to run in a parallelized mode. The result, an expanded spatial model, will be extremely useful to ecologists, geographers, geologists, and others modeling large systems.
"Larry Smarr calls it a superb knowledge capturing device, and I agree with him. Jim Westervelt likens it to the game of chess: rules that are very easy to learn coupled with infinitely complex strategies."
Hannon uses STELLA software in other classes he teaches: Biological Modeling, Modeling Resource Scarcity, and an undergraduate honors class called Dynamic Modeling. Students in the classes are from a variety of disciplines at UIUC within engineering, business, agriculture, health sciences, natural sciences, physical sciences, and medicine.
According to its developers, STELLA II is built on the systems approach--sometimes called Systems Thinking--to problem solving. Using this methodology, the emphasis is placed on broad viewpoints--or the "big picture" view--so that interrelationships and interconnectivity are the focus rather than a collection of complex variables. Its proponents call this process a twenty-first century attitude towards problem solving.
STELLA II is a "completely rewritten and extended version of STELLA for Education," introduced in 1987, according to its developers. The software is used in a variety of disciplines in natural science, social science, and humanities at the secondary through postgraduate levels.
STELLA II was developed at High Performance Systems (HPS), which was founded in 1985. The company spun off from founder Barry Richmond's consulting practice. When he began HPS, Richmond was a professor of engineering and business at Dartmouth College, Hanover, NH.
For more information about STELLA II, contact High Performance Systems, 45 Lyme Road, Suite 300, Hanover, NH 03755; Applelink (electronic mail): X0858; phone (603) 643- 9636; FAX: (603) 643-9502.
__________________________________________________________________
access / October-December 1992 / NCSA / pubs@ncsa.uiuc.edu